JPH0490265A - Picture processor - Google Patents

Abstract

PURPOSE:To attain both a high compression rate and high picture quality by automatically applying image area separation and compressing data depending on the compression system suitable for each area. CONSTITUTION:An image area separate section 3 divides a multi-valued picture area and a binary picture area of an input picture and writes 0 to a location of a picture element of the binary picture area and 1 to a picture location of the multi-valued picture area in a separate result memory 4. A binary compression/decoding section 4 reads the separate result data from the separate result memory 4 and forcibly brings picture elements other than those of the separated areas to white level picture elements, the result is binarizing-processed and the data is compressed by the compression system of the binary picture and stored in a binary compression data memory 5. Moreover, picture elements other than those of the separated areas as the multi-valued picture area are brought forcibly to white level picture elements, the result is compressed by the compression system of the multi-valued picture and the obtained multi-value compressed data is stored in the multi-valued compression data memory 6.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image processing apparatus that compresses and stores, transmits, or reproduces images such as documents.

[Prior art] In image file systems and still image transmission systems,
When targeting an image containing halftones such as a photograph, the original image is read as multi-valued data by a scanner, undergoes pseudo-halftone processing such as dithering, and then performs binary image compression such as Modified Huffman (MH) encoding. Compressing data through encoding, recording it on an information storage medium or transmitting it to the outside, or decoding the compressed data received from the information storage medium or the line and reproducing it on a binary printer etc. There are many. However, the quality of reproduced halftones is worse than that of the original image, and the resolution of character images is considerably degraded due to the effect of pseudo halftone processing. Also, the data compression rate is insufficient.

In order to solve the second problem, various methods for processing images by image area have been studied. An example is shown below.

Convert original image data inputted by a conventional multi-value scanner into binary data and multi-value data with a reduced number of gradations, or into pseudo-halftone data using a dither method, and then convert it into pre-specified areas. 2 of the manuscript according to the location information
In the value image area, binary data is selected and output, and in the halftone area of the document, multivalue data or pseudo halftone data is selected and output (Japanese Patent Laid-Open No. 63-59264, Japanese Patent Laid-Open No. 53
-142964).

Conventional technology ■: A method suitable for binary images in which a photographic area is extracted as a rectangle from an input image, the photographic area of the input document is replaced with white or black pixels line by line, and then binarized and MR encoding is performed. On the other hand, the photographic area is formatted and then compressed using a method suitable for multivalued images, such as hector quantization or transform encoding, and the two types of compressed data obtained in this way are to store or transmit.

When reproducing an image, the compressed data of the photo area and the other compressed data are decoded using methods suitable for each and combined (Japanese Patent Laid-Open No. 144778/1999).

Prior art (2) For an image input as multivalued data, it is determined whether it is halftone or binary in units of predetermined blocks. Halftone blocks are converted to the average density value within the block, and binary blocks are converted to a bit pattern that corresponds to the density pattern within the block, thereby compressing the amount of data and storing or transmitting it. Kaisho 6
3-191475).

Conventional technology ■: Binary image data and multi-value image data of a document are stored in separate memories, coordinate information of the multi-value image area is also stored, and each memory is stored according to the coordinate area of the multi-value image area. is selectively read out, and the binary image data is also converted to a multi-value level to synthesize an image containing a mixture of binary image areas and multi-value image areas, which is then D/A converted and converted into C
4 (Output to T-display, etc.)
No. 8968).

[Problems to be Solved by the Invention] Conventional techniques (1) and (2) require specifying a halftone or multivalued image area in advance, and therefore cannot be applied to applications that process arbitrary images. Further, when pseudo halftone processing using the dither method is performed, deterioration of image quality in the halftone region is unavoidable. Note that the relevant patent publication does not mention data compression.

Conventional technology ■ is superior in terms of both reproduction image quality and data compression rate, provided that the accuracy of rectangular extraction of the photo area and its shaping process, as well as the accuracy of compositing the photo area and other areas, are sufficiently high. You can expect good results. However, it is not easy to extract and reshape photographic areas with high precision automatically for a variety of input images, and the processing time tends to increase, and the quality of the reproduced image can be significantly affected by a small processing error. There is a risk that it will get worse.

Although the conventional technique (2) is easy to process, it tends to cause deterioration in the quality of halftones, and there is also a limit to the improvement in the data compression rate.

The purpose of the present invention is to
An object of the present invention is to provide an image processing device capable of efficient data compression and high-quality image reproduction through relatively simple processing for any image containing a mixture of value regions.

[Means for Solving the Problems] The invention described in claim (1) provides an image area separation means for separating a binary image area and a multivalued image area of an input image input as multivalued data; Binary image compression means for converting the input image into binary compressed data as an image in which areas other than the binary image area are white according to the image area separation result by the image area separation unit; a multivalued image compression means for converting the image into multivalued compressed data, recording the binary compressed data by the binary image compression means and the multivalued compressed data by the multivalued image compression means on an information recording medium, or The invention as set forth in claim (2) is characterized in that the image processing apparatus according to the invention as set forth in claim (1) includes a means for outputting the binary compressed data to the outside; Multi-value image decoding means for decoding compressed value data, converting the binary image data decoded by the binary image decoding means into a multi-value level and combining it with the multi-value image data decoded by the multi-value image decoding means. The invention as set forth in claim (3) is characterized in that the invention comprises an image compositing means for performing multi-level recording of composite image data by the image compositing means, and an image means for performing multi-value recording of composite image data by the image compositing means. The apparatus includes a determination means for determining whether the input image is a binary image or a binary and multivalue mixed image based on the characteristics of the region determined to be a multivalued image region by the image area separation means, The invention as set forth in claim (4) is characterized in that an input image determined to be a binary image by the means is processed as a binary image. Binary compressed data and multi-value compressed data of an image in which areas other than the multi-value image area of the input image are white,
A binary image decoding means input from an information storage medium or a communication line and decoding the binary compressed data, a multi-value image decoding means decoding the multi-value image data, and a binary image decoded by the binary image decoding means. It has an image composition means for converting image data into a multi-value level and composites it with the multi-value image data decoded by the multi-value image decoding means, and an image recording means for performing multi-value recording of the composite image data by the image composition means. It is characterized by

[Function] According to the invention described in claim (1), the image of various input images in which multivalued image areas (for example, halftone areas such as photographs) and binary image areas (for example, characters, etc.) coexist can be adjusted. Even if the area is not specified in advance, image area separation is automatically performed and 2
By efficiently compressing the data of a value image area using a suitable binary image compression method with less image quality deterioration, and by compressing the data of a multi-value image area using a suitable multi-value image compression method with less image quality deterioration. , it is possible to simultaneously achieve high compression ratio and high image quality. Therefore, troublesome processing such as highly accurate image area rectangle extraction and its shaping is not necessary, and the processing content is simple and high-speed processing is easy.

In addition, binary compressed data is data that has been compressed with areas other than the binary image area of the input image being white, and multilevel compressed data is also data that has been compressed with areas other than the multilevel image area of the input image being white. According to the invention described in item (2) or (4), it is possible to restore original image data by simply adding the respective decoded data, and it is possible to reproduce it with high image quality by recording it in multiple values.

According to the invention described in claim (3), since an input image consisting only of binary images is processed as a binary image, unnecessary recording, transmission, and decoding of multi-value compressed data can be eliminated.

[Example] FIG. 1 is a block diagram showing one embodiment of the present invention.

In FIG. 1, 1 is a multi-level image scanner that reads a document, 2 is an input/output image memo I for storing input image data and restored images from the image scanner, and 3 is a binary image area of the input image. An image area separation unit for separating multivalued image areas; 4 is a separation result memory that stores the results of image area separation; 5 is a binary compressed data memo 1 that stores binary compressed data;
Halo is a multi-value compressed data memory that stores multi-value compressed data. 7 is a binary compression/decoding section that performs binary image compression processing and decoding processing, and 8 is a multi-value compression/decoding section that performs multi-value image compression processing and decoding processing. Reference numeral 9 denotes a determination unit that uses the function of the multi-value compression/decoding unit 8 to determine whether the input image is a binary image or a binary multi-value mixed image; 10 represents each memory 5;
. 11 is a transmission unit that transmits compressed data of an input image to an external device through a communication line or receives compressed data from an external device;
12 is an image synthesis unit that synthesizes decoded data of compressed data; 13 is a printer that performs multi-value recording of multi-value image data;
Reference numeral 4 denotes a control unit that controls the operation of each unit and data transfer. There is also an operation section where the user can give commands, etc.
Not shown.

The operation will be explained below.

A document is read by an image scanner 1, and the document image is written into an input/output image memory 2 as, for example, 8-bit multi-value data. The image area separation unit 3 reads the multivalued data of the input image from the input/output image memory 2, separates the multivalued image area and the binary image area of the input image, and stores the data in the separation result memory 4, for example, the binary image area. O1 at the pixel position of
Write 1 at the image position in the multivalued image area.

This region separation can be performed by various methods, but for example, as described in detail in the specification and drawings of Japanese Patent Application No. 1,113,242 filed by the same applicant, white pixel outlines of characters and connections are used. Focusing on the fact that black pixels exist simultaneously at a density higher than a certain level, we first classify each pixel into black pixels and white pixels from the multi-value data of the input image, and then create a black pixel that is connected from the classified black pixel group and white pixel group. Pixels and white pixels are respectively detected, and the detected connected black pixels and connected white pixels both exceed a certain number. 2. The pixels existing in the vicinity are detected as temporary character pixels, and the temporary character pixels are fixed in the vicinity of the pixel of interest. When there are more than the number of pixels, a method can be used in which the pixel of interest is determined to be a pixel of a character portion.

With such image area separation, for example, in the case of an input image in which a text portion and a photo portion coexist as shown in Fig. 2, the text portion and the adjacent image are separated as binary image regions as shown in Fig. 3. , the other photographic areas and white background areas are separated as multivalued image areas.

The binary compression/decoding unit 7 reads the multivalued data of the input image from the input/output image memory 2 and performs binarization processing using a predetermined binarization threshold, but at the same time reads the separation result data from the separation result memory 4. , pixels other than the area separated as a binary image area of the input image are forced to be white pixels.

In other words, one page of input image is binarized as one page image with white except for the binary image area. Then, this binary data is compressed according to the compression method number of the binary image, and the obtained binary compressed data is stored in the binary compressed data memory 5. In this embodiment, it is assumed that the MH encoding method, which is standardly adopted in facsimile, is used as the compression method. According to this encoding method, it is possible to avoid resolution degradation that occurs in the dither method.

In addition, the multilevel compression/decoding unit 8 reads multilevel data of the input image from the input/output image memory 2, simultaneously reads separation result data from the separation result memory 4, and performs binary compression/decoding.
Similar to the decoding unit 7, the multi-value data of the input image is compressed using the multi-value image compression method, with pixels other than the area separated as the multi-value image area being white, and the obtained multi-value compressed data is The data is stored in the multivalued compressed data memory 6. In this example, ADC (ADCT), which is a standard still image compression method, is used.
Adaptive discrete cosine transform, A adaptive D 1s
crer, eCosine Transfer'orm) is adopted as the compression method. According to this compression method, it is possible to avoid image quality deterioration that occurs in the dither method.

In this way, one page's worth of binary compressed data and one page's worth of multilevel compressed data are obtained for one page of input image, but both compressed data are stored in the optical disk device 10.
and recorded on an optical disc according to a predetermined format. If it is not necessary to save both compressed data, it is also possible to immediately transmit them from the transmission unit 11 to an external device through a communication line.

In addition, if you use an external device with a similar compression processing system,
It is also possible to take in the value compressed data and the multi-value compressed data by the transmission unit 11 through a communication line, store them in the memories 5 and 6, and then record them in the optical disk device 10.

In the case of printing out an image, the compressed data of the necessary image is retrieved from the optical disk device 10 and reproduced from the optical disk, or it is imported from an external device through the transmission section 11, and the binary compressed data is stored in the binary compressed data memory 5. The multi-value compressed data are respectively written to the multi-value compressed data memory 6.

The multi-level compression/decoding section 8 and the binary compression/decoding section 7 read and decode the multi-level compressed data and the binary compressed data from the multi-level compressed data memory 6 and the binary compressed data memory 5, respectively, and produce the decoded multi-level image. The image synthesis unit 1 alternately processes data and binary image data in predetermined processing units (for example, 8 lines).
Transfer to 2.

The image synthesis section 12 has an image synthesis memory for processing units (for example, 8 lines) internally, and first writes the transferred decoded multi-valued image data for the processing units into pixel-corresponding positions in the image synthesis memory. Subsequently, the decoded binary image data transferred is converted into multilevel data, and then written into the pixel-corresponding position of the image synthesis memory.

For example, if the decoded multivalued image data is 8-bit data, 256 gradations from O for white to 255 for black can be expressed. On the other hand, decoded binary image data expresses white as 0 and black as 1. Therefore, in the image synthesis section 12, as shown in FIG.
For the pixel (black), 255 is written to the corresponding position in the image synthesis memory.

Through such processing, additive synthesis of decoded multivalued image data and binary image data is achieved. The reason why such a simple compositing process is possible is that when compressing data, multi-value image compression is performed with areas other than the multi-value image area being white.
This is because binary image compression is performed with areas other than the value image area being white, and there is no overlap between the multivalue image area and the binary image area.

The multivalued image data for the processing unit synthesized on the image synthesis memory is transferred to the input/output image memory 2, and the image synthesis process continues for the next processing unit.

When one page's worth of multivalued image data is stored in the input/output image memory 2, it is transferred to the multivalued printer 13 and printed out in multivalued form.

As explained above, according to the present invention, one page of input image is a total of two pages of images, one page with white areas other than the multilevel image area, and one page with white areas other than the binary image area. Data is compressed using each compression method. In either data compression, the area of the white background part of the target image increases, but the compression rate of the white background part is high, and the compression method that is most suitable for the image area is applied to each data compression, and the rectangular extraction of the multivalued image area is Unlike methods that compress data by performing data compression, there is no need to create data representing multivalued image areas, so a high overall data compression rate can be achieved, image quality deterioration can be minimized, and processing is simple. It is easy to increase the speed.

In addition, for image restoration, all that is required is to decode each compressed data, add and combine them, and the process is simple and fast, and a high-quality restored image can be obtained.

Now, it is thought that most of the morphometric document images consist of only binary images, and there are few binary and multi-value mixed images. Therefore, if input images are indiscriminately processed as both binary images and multi-valued images as described above, a large number of documents, most of which consist only of binary images, will be continuously scanned by the scanner 1. This is not necessarily efficient when the data is read and processed by a computer. For document images consisting only of binary images, it is clearly more efficient to process them as binary images (recording on an optical disk, transmitting, restoring).

Therefore, according to the present invention, it is determined whether the input image is a binary and multivalued mixed image, and if the input image is determined to be a binary image, it is processed as a binary image. can be controlled.

In the embodiment shown in FIG. 1, a determination section 9 is provided for such determination. Regarding this judgment section 9, the fifth
This will be explained using figures.

The multi-value compression/decoding unit 8 uses ADC as a multi-value image compression method.
In order to utilize T, an ADCT section 16 as shown in FIG.
There are 6 generals. This ADCT department 16 division general 6-dimensional DCT
(Discrete cosine transformation) unit 17 transforms the input image data into 8×8
The configuration is such that pixels are cosine transformed, the transformed values are quantized by a quantizer 18, and then encoded by a Huffman encoder 19.

In two-dimensional cosine transformation, one DC (
It is known that a direct current) component and 63 AC (alternating current) components can be obtained. DC component is +1025 to -1024
up to 11 bits, and the upper 8 bits of +
127 to -128 means the average value of luminance within the 8×8 block.

In Uni, -127 indicates the white level and +128 indicates the black level.

Normally, in the midtone area (picture area) made up of halftone dots in photographs and printed matter, areas with a certain high density gather to form the pattern part, whereas in the background area excluding text areas, the average concentration is very low. Therefore, the density of the area excluding the character area can be used as a characteristic for determination.

The determining unit 9 makes such a determination, and the simplest method is as follows. That is, the two-dimensional DCT section 17
The DC component obtained for the 8×8 block is compared with the threshold value, and if the DC components of all blocks in the area determined as the multilevel image area by image area separation are lower than the threshold value, (DC
If there is no block whose component exceeds the threshold value), it is determined that there is no picture part in the area designated as a multi-valued image area, that is, the input image is a binary image.

However, according to such a simple judgment method, even if an input image consists of only a few characters, if an area with a certain level of high density occurs due to dirt on the background, this will be mistaken for a pattern, and it will be interpreted as a binary and multi-value mixed image. There is a risk of misjudgment. Therefore,
More preferably, the determination unit 9 performs the determination in the following manner.

In other words, if there is a picture area, the density should exceed the threshold in at least some continuous area.
In the area converted into a multivalued image area by image area separation, DC
The number of blocks whose components exceed the threshold is counted, and when the number exceeds a specified value (for example, 5 or more blocks in a row), it is determined that a pattern exists, that is, the input image is a binary and multivalued mixed image. It is determined that there is. According to this determination method, it is difficult to mistake minute dirt on the background for the pattern, so
More reliable judgment becomes possible.

Note that the method of using the DC component of the two-dimensional DCT for determination is the two-dimensional DCT in the multi-value compression/decoding unit 8 using ADCT.
There is an advantage that the function of section 17 can be used. However, the same determination may be made by directly determining the average density of a minute portion (nXn) of a region defined as a multi-value image region from the multi-value data of the input image.

When the determination unit 9 determines that the input image is a binary image as described above, the control unit 14 converts the input image into a binary image.
Process it as a value image. That is, the contents of the multilevel compressed data memory 5 are invalidated, and only the contents of the binary compressed data memory 5 are recorded in the optical disk device 10 or transmitted by the transmission section 11. In the case of image reproduction, only the binary compressed data is restored and sent to the image composition section 8.

Note that when handling a color image, similar image area separation and determination may be performed focusing on the luminance (a degree) component, and each color component may be processed independently. Furthermore, image area separation may be performed to separate into three or more types of areas, including halftone areas.

[Effects of the Invention] As explained in detail above, according to the invention recited in claim (1), multivalued image areas (for example, intermediate r
For various RokunoJ images that contain a mixture of I4 area) and binary image area (for example, characters), even if the image area is not specified, image area separation is automatically performed to separate the binary image area. A binary image compression method (S
By compressing data more efficiently and compressing multi-value image areas using a multi-value image compression method suitable for that purpose and with less image quality deterioration, it is possible to simultaneously achieve high compression ratios and high image quality. The content is simple and high-speed processing is easy. In addition, binary compressed data is data that has been compressed with areas other than the binary image area of the input image being white, and multilevel compressed data is also data that has been compressed with areas other than the multilevel image area of the input image being white. According to the invention described in item (2) or (4), it is possible to restore the original image data by simply adding the respective decoded data, and by recording the data in multiple values, it is possible to reproduce a high-quality image. be. Furthermore, according to the invention set forth in claim (3), an input image consisting only of binary images is processed as a binary image, and efficient processing is achieved by eliminating unnecessary recording, transmission, and decoding of multilevel compressed data. It is possible.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram of an example of an input image, FIG. 3 is an explanatory diagram of the result of separating character areas of the input image shown in FIG. FIG. 4 is a flowchart illustrating image synthesis of decoded binary image data, and FIG. 5 is a block diagram illustrating the configuration of the ADCT unit and the operation of the determination unit in the multilevel compression/decoding unit. l...Multilevel image scanner 2...Input/output image memory, 3...Image area separation unit,
4... Separation result memory, 5... Binary compressed data memory, 6... Multilevel compressed data memory, 7... Binary compression/decoding section, 8... Multilevel compression/decoding section, 9. ...Judgment section, 10
...Optical disk device, 11...Transmission section, 12...
・Image synthesis unit, 13...Multi-value output printer, 14
...control section, 16...ADCT section, 17...
Two-dimensional DCT unit, 18... Quantization unit, 19... Huffman encoding unit. Figure 2

Claims (4)

[Claims] (1) An image area separation means for separating a binary image area and a multivalued image area of an input image input as multivalued data, and according to the image area separation result by the image area separation means, the input image is separated from the binary image area. a binary image compression means for converting an input image into binary compressed data as a white image, and a multi-value image compression means for converting an input image into multi-value compressed data as a white image except for the multi-value image area; An image processing apparatus comprising means for recording binary compressed data produced by a value image compression means and multivalued compressed data produced by the multivalued image compression means on an information storage medium or outputting them to the outside. (2) Binary image decoding means for decoding binary compressed data, multi-value image decoding means for decoding multi-value compressed data, and converting the binary image data decoded by the binary image decoding means into multi-value levels. Claim (1) characterized in that the method further comprises an image composing means for composing the multi-valued image data decoded by the multi-valued image decoding means, and an image recording means for performing multi-valued recording of the composite image data by the image composing means. ) image processing device. (3) A determination means for determining whether the input image is a binary image or a binary and multivalue mixed image based on the characteristics of the region determined to be a multivalued image region by the image area separation means; 2. The image processing apparatus according to claim 1, wherein the input image determined to be a binary image by the means is processed as a binary image. (4) Image 2 with white areas other than the binary image area of the input image
Binary image decoding means for inputting value compressed data and multi-value compressed data of an image in which areas other than the multi-value image area of the input image are white from an information recording medium or a communication line and decoding the binary compressed data; Multi-value image decoding means for decoding value image data, converting the binary image data decoded by the binary image decoding means to a multi-value level and combining it with the multi-value image data decoded by the multi-value image decoding means. 1. An image processing apparatus comprising: an image synthesizing means for performing multi-level recording of synthesized image data by the image synthesizing means;